HTIX: Hate Speech Detection Using Vector Search and Generative AI
Abstrak
E-commerce applications on the Google Play Store have become a major platform for online transactions and user feedback through reviews. However, the increasing volume of user-generated content has led to the emergence of hate speech, which can harm sellers, reduce platform credibility, and negatively affect user experience. Manual moderation of such content is inefficient and prone to inconsistency, highlighting the need for an automated detection system. To address this problem, this study proposes a web-based system named HTIX for automatically detecting and classifying hate speech in e-commerce application reviews.The main contribution of this research lies in integrating vector search database technology with generative AI to enable semantic-based hate speech detection in real-world e-commerce review data.The proposed system applies vector embeddings to represent review texts and performs semantic similarity searches using a vector search database. Generative AI is then utilized to assist in analyzing and classifying the reviews into hate speech and non-hate speech categories. Experimental results showed that HTIX was able to classify reviews collected from e-commerce applications on the Google Play Store into Hate Speech and Non-Hate Speech categories. From 100 analyzed reviews, the system identified 6 reviews as Hate Speech and 94 reviews as Non-Hate Speech. The generated confidence scores were displayed for each prediction result to provide additional information regarding the model’s classification certainty. Visualization outputs in the form of pie charts and prediction tables presented the distribution of review classifications and detailed prediction results. The pie chart illustrated that the majority of analyzed reviews were categorized as Non-Hate Speech, while the prediction table displayed review texts, classification labels, and confidence percentages for each analyzed review. These results demonstrate the capability of the proposed system to support automated hate speech analysis and provide interpretable visualization outputs for review moderation.
Identifier
- DOI: https://doi.org/10.31154/cogito.v12i1.1075.69-80
- OpenAlex: https://openalex.org/W7202186210